MIDAS vs. Mixed-frequency VAR: Nowcasting GDP in the Euro Area

摘要: Abstract This paper compares the mixed-data sampling (MIDAS) and mixed-frequency VAR (MF-VAR) approaches to model specification in presence of data, e.g. monthly quarterly series. MIDAS leads parsimonious models which are based on exponential lag polynomials for coefficients, whereas MF-VAR does not restrict dynamics can therefore suffer from curse dimensionality. However, if restrictions imposed by too stringent, perform better. Hence, it is difficult rank a priori, their relative rankings better evaluated empirically. In this paper, we compare performances case relevant policy making, namely nowcasting forecasting GDP growth euro area monthly basis, using set about 20 indicators. It turns out that two more complements than substitutes, since tends horizons up four five months, performs longer horizons, nine months.